This is a single-center, phase I/II, individual dose escalation study of the effect of cytoreductive chemotherapy combined with aggressive antiretroviral therapy on lymph node HIV DNA (viral load) in ten treatment-naive HIV-infected male and female subjects with CD4 T-cell counts >300 cells/mm3. The primary endpoints are (1) the difference in the change in integrated viral DNA in lymph nodes from baseline to week +18 between the two treatment groups, and (2) the safety of the antiretroviral therapy plus cytoreductive chemotherapy as measured by the absence of any significant or serious toxicities. Eligible subjects will enter Step 1 and will receive up to 16 weeks initiation of aggressive antiretroviral therapy (HAART; highly active antiretroviral therapy) consisting of nelfinavir 750 mg t.i.d. + d4T 40 mg b.i.d. for subjects who weigh 360 kg or as 30 mg b.i.d. for subjects who weigh >=60 kg + 3TC 150 mg b.i.d. Subjects who demonstrate 2 consecutive plasma HIV RNA determinations <50 copies/mL by the Roche Ultrasensitive assay, in a four week period, between week 4 to week 24, after initiation of HAART, will be randomized to Step 2 to receive HAART alone (ARM A) or HAART plus cytoreductive chemotherapy (cyclophosphamide) ARM B. The cytoreductive chemotherapy will be given as three single doses over a 12-week period and will be dose escalated two times. All three doses will be administered in the Duke University General Clinical Research Center (GCRC). The cytoreductive chemotherapy will be administered as a dose escalation with one hour IV infusions of cyclophosphamide starting at 750 mg/m2, followed by 1.2 g/m2 and 1.8 g/m2 at 6 week intervals (3.75 g/m2 total dose). Each of those dosages are below the maximum tolerated dose (MTD) for single dose cyclophosphamide which is 4 g/m2. On the week of randomization to Step 2, all subjects will have baseline evaluations to include safety labs, advanced flow cytometry, plasma HIV RNA by the Roche Ultrasensitive assay (batched), PBMC HIV RNA and DNA, HIV quantitative co-culture, latent T-cell assay, TCR V-beta by flow cytometry (if available), CDR3 polymorphism, and an excisional lymph node biopsy. Specimens of all types will be archived for future analyses. Lymph node biopsy for quantitative HIV RNA/DNA measurement will be repeated at week +20 after randomization to Step 2 or six weeks after the last cycle of cyclophosphamide and will be correlated with PBMCs collected at similar time points to determine whether the number of latently HIV-infected cells in peripheral blood correlates with that of lymphoid tissues. Pharmacokinetic (PK) studies will be performed in the GCRC to determine potential PK interactions between cyclophosphamide and the antiretroviral agents. On each PK day, a light to moderate size breakfast with moderate fat content should be provided (nelfinavir AUCs were 2- to 3-fold higher under fed conditions compared to fasting). PK day dose of nelfinavir, d4T and 3TC may be taken together after 00:00 time sample before cyclophosphamide is administered. The cyclophosphamide administration should begin approximately 30 minutes after the nelfinavir is administered. Subjects will be followed for a total of 52 weeks after randomization to Step 2 (Cyclophosphamide +HAART or HAART alone). Cerebrospinal fluid will be obtained (optional) at week +52 to determine the amount of HIV RNA and DNA present. Chest CT of the thymus will be performed at week of randomization and week +52.